Low-level feature extraction is the first step in any image analysis procedure 
and is essential for the performance of stereo vision and object recognition 
systems. Research concerning the detection of corners, blobs and circular or 
point like features is particularly rich and many procedures have been proposed 
in the literature. In this paper, several frequently used methods and some novel 
ideas are tested and compared. We measure the performance of the detectors under 
the criteria of their detection and repeatability rate as well as the 
localization accuracy. We present a short review of the major interest point 
detectors, propose some improvements and describe the experimental setup used 
for our comparison. Finally, we determine which detector leads to the best 
results and show that it satisfies the criteria specified above.